The Information Machine

Google's new Gemma 4 12B model is designed to run on any laptop with 16GB of RAM

Ars Technica AI · Ryan Whitwam · 2026-06-03

Google released Gemma 4 12B, a new open-weight model requiring only 16GB of RAM that fills the capability gap between the Gemma 4 family's mobile-optimized and high-end variants, and which Google claims performs nearly as well as the 26B MoE model on benchmarks.

Open original ↗

Appears in

Extraction

Topics: open-weight-modelson-device-aigoogle-gemmalocal-inference

Claims

  • Gemma 4 12B requires only 16GB of system RAM or VRAM, making it runnable on many consumer laptops.
  • Google claims Gemma 4 12B is nearly as capable as Gemma 4 26B MoE on benchmarks while requiring roughly half the memory footprint.
  • The new 12B model fills a gap in the Gemma 4 lineup between the mobile-optimized E2B/E4B models and the larger 26B MoE and 31B Dense models.
  • Gemma 4 models use an Apache 2.0 license, introduced with the April 2026 release.

Key quotes

Google says Gemma 4 12B is unique in that it can run on many consumer laptops without sacrificing quality.